Terra Truce: The Project Story

Inspiration

Terra Truce was born from a simple but frustrating insight: real estate—one of the world’s most valuable asset classes—remains deeply opaque. While equity and crypto investors have access to real-time analytics, simulations, and automated risk tools, property investors still rely on fragmented data, manual legal reviews, and intuition-driven decisions.

We set out to create a single pane of glass for real estate intelligence—one that turns raw property, legal, and environmental data into clear, actionable investment insights. Terra Truce represents a “truce” between market uncertainty and investor confidence.

What it does

Terra Truce is a next-generation property intelligence platform designed to de-risk real estate investments using agentic AI and machine learning. It provides:

• Real-Time Location Risk Analysis
  Live assessment of environmental hazards (flood zones, air quality) and social infrastructure factors (crime rates, transport access) via real-time web intelligence.

• Legal Red-Flag Detection
  AI-powered OCR and NLP scan property deeds and legal documents to identify encumbrances, disputes, and regulatory non-compliance.

• Financial Forecasting
  Long-term (up to 30-year) ROI and appreciation projections using trained XGBoost models grounded in real economic indicators.

• Smart Portfolio Management
  A real-time, multi-currency dashboard that tracks asset performance, risk exposure, and long-term value health.

How we built it

Terra Truce is engineered as a high-availability, microservices-based system designed for intensive AI orchestration:

• Core Engine
  A Rust-based API gateway and MCP orchestrator provide memory safety, low latency, and extreme concurrency for multi-agent AI workflows.

• Intelligence Layer
  Dockerized workers handle specialized tasks:

  • Python services for XGBoost financial modeling
    
  • Node.js services for data synthesis and enrichment

• AI Orchestration
  Model Context Protocol (MCP) enables coordinated reasoning across multiple AI systems, including:

  • Gemini 3 Pro for document reasoning
    
  • Perplexity for real-time risk intelligence

• Frontend Interface
  Built with React 19, Vite, and Tailwind CSS, enhanced with Framer Motion to deliver a fluid, glassmorphic user experience.

Challenges we ran into

• Concurrent AI Synthesis
  Combining outputs from multiple AI agents into a unified real-time risk score initially caused significant latency. We solved this by:

  • Migrating orchestration to Rust (Axum) for asynchronous efficiency
    
  • Introducing Redis-backed caching for frequently accessed location and model data

• Privacy in OCR Pipelines
  Legal documents contain highly sensitive data. We implemented a custom PII redaction pipeline to ensure all personally identifiable information was stripped before any cloud-based LLM processing.

Accomplishments that we're proud of

• Seamlessly combining static legal analysis with dynamic, real-time risk factors in a single platform
  
• Accurately projecting long-term ROI while detecting legal red flags in complex, multi-page property documents
  
• Successfully implementing the Model Context Protocol (MCP) to enable standardized communication between AI agents

What we learned

This project was a deep dive into agentic AI, financial engineering, and systems design. We learned how to balance the probabilistic reasoning of LLMs with the mathematical rigor of traditional ML models like XGBoost.

We also gained hands-on experience applying real-world financial formulas, ensuring projections reflected true investor value by accounting for inflation:

• Future Value:
  FV = PV × (1 + r)ⁿ

• Real ROI:
  Real ROI = (1 + Nominal ROI) / (1 + Inflation Rate) − 1

What's next for Terra Truce

Our roadmap includes:

• Climate Change Predictive Modeling
  Forecasting how properties may be affected 50+ years into the future.

• Blockchain-Based Title Verification
  Automating ownership and title verification to further streamline legal due diligence.

• Open-Sourcing the MCP Orchestrator
  Releasing our Rust-based MCP orchestration layer to empower the broader developer community.

Built With

• javascript
• mcp
• react
• rust
• supabase